Autosubmit server

Autosubmit is designed to run on a central server, which allows easy sharing of experiment configurations and a common experiment database. At KIT, such a server is provided and this page contains info on how to access the server and how to use it.

Important

The server is only available for KIT users. Other users should install their own (local) instance of Autosubmit Autosubmit installation.

Connection details

The server can be access via ssh at the domain name auto-icon.scc.kit.edu. If you want to view the plots created by Autosubmit, X forwarding is required (with the -X flag or the config below). Login with your KIT account, i.e. your id (ab1234) and the corresponding password. Furthermore (after the first login), you can also access the server with an encrypted SSH key. You might want to consider adding the server to your .ssh/config file, e.g.:

Host autosubmit
        HostName auto-icon.scc.kit.edu
        User <username>
        PasswordAuthentication yes
        PreferredAuthentications keyboard-interactive,password
        GSSAPIAuthentication no
        PubkeyAuthentication no
        IdentitiesOnly no
        ForwardX11 yes

Connecting with VS Code

A VS Code server is installed on the server. Thus, remote development can easily be set up following the official guide.

Connecting to the Autosubmit GUI

Important

The Autosubmit GUI has to be installed and set up separately. At KIT it is also installed and hosted on the Autosubmit server.

The Autosubmit GUI (for monitoring experiments) is available at https://auto-icon.scc.kit.edu:44321. You will have to log in with a Github account.

If the connection is refused, ensure you are connected to the KIT network (directly or via VPN).

Note: currently, the server is using a self-signed certificate, thus, you will have to accept the risk and continue.

SSH connections to other systems

SSH connections from this server to HPC systems as well as to the DKRZ Gitlab have to be set up on this server. It is recommended to use a separate SSH key for each site. Follow the instructions on the setup page for a full setup.

Autosubmit versions on the server

On the server, multiple Autosubmit versions are installed. Switching between these is possible with the asv tool:

$ asv -h
Usage: asv VERSION
       asv load VERSION
       asv unload
       asv list
       asv help|-h|--help
       'load' load the autosubmit version denoted by VERSION if existing.
          'asv VERSION' is a convenient short cut for asv load VERSION.
       'unload' unloads the current version.
       'list' lists all available versions.
       'help', '-h' or '--help' prints this usage info.

You can check the current Autosubmit version with autosubmit --version. To update your experiment from one version of Autosubmit to another, you can run autosubmit updateversion <expid>. Per default, the ‘stable’ version of Autosubmit is loaded.

Internally, asv just switches between different Python virtual environments. Keep this in mind in case you need (for whatever purposes) an own venv.

The purpose of the server

The server has the purpose to hold and share all the experiment configurations. It is there to administer all your experiments, create and change the configuration and monitor that everything runs smoothly.

Warning

It is not meant to serve as a workstation for analyzing your results, storing your input or output data and so on. The latter is still meant to reside on whichever platform you run your experiments at.

Warning

Sometimes ICON logs can get really large, esp. when writing debug output (~GB). As all log files are transferred to the server, please take care in such cases and delete the logs at least from the server.